Training in progress, step 100, checkpoint

checkpoint-100/adapter_config.json

@@ -20,10 +20,10 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "o_proj",
     "q_proj",
-    "k_proj",
-    "v_proj"
+    "v_proj",
+    "o_proj",
+    "k_proj"
   ],
   "task_type": "CAUSAL_LM",
   "use_dora": false,

checkpoint-100/adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8204c22c3e3d5a41ea66ebf806f58168719a2fd235638d163fd97c5b42c39ec0
+oid sha256:0a09efd490b1230a03de342cc81ebd092ee71bdf105cf41f55fd564ccfb7fe67
 size 54560368

checkpoint-100/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d555711955935cd1a5e4ead0e6392f6e3236611ae71a091b7218017f9a9cd0b6
+oid sha256:31ec931eae12d877a7421bb24a3cd866d5cb82e68cb89575eac37dd6eb7bf739
 size 109267450

checkpoint-100/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d18870198c31c5821df8b9e9d648f47112d84cad0aef4ef7508fe2f35542f854
+oid sha256:99cdf514d195ace594a247b30bc61c0b0ccb08b6431f56f75e4d6b9fc7392b92
 size 1064

checkpoint-100/trainer_state.json

@@ -1,5 +1,5 @@
 {
-  "best_metric": 0.4733425974845886,
+  "best_metric": 0.47333332896232605,
   "best_model_checkpoint": "./zephyr/05-04-24-Weni-WeniGPT-Agents-Zephyr-1.0.1-KTO_testing kto dataset during training-3_max_steps-145_batch_16_2024-04-05_ppid_9/checkpoint-100",
   "epoch": 0.684931506849315,
   "eval_steps": 50,
@@ -10,95 +10,95 @@
   "log_history": [
     {
       "epoch": 0.14,
-      "grad_norm": 8.764093399047852,
-      "kl": 11.000414848327637,
+      "grad_norm": 4.650723457336426,
+      "kl": 4.865365505218506,
       "learning_rate": 0.00018285714285714286,
-      "logps/chosen": -320.3092041015625,
-      "logps/rejected": -296.3908386230469,
-      "loss": 0.432,
-      "rewards/chosen": 1.4254474639892578,
-      "rewards/margins": 1.3692480325698853,
-      "rewards/rejected": 0.0951162651181221,
+      "logps/chosen": -310.4614562988281,
+      "logps/rejected": -364.2945556640625,
+      "loss": 0.45,
+      "rewards/chosen": -2.890953779220581,
+      "rewards/margins": 3.259232521057129,
+      "rewards/rejected": -5.586672306060791,
       "step": 20
     },
     {
       "epoch": 0.27,
-      "grad_norm": 5.878363609313965,
-      "kl": 1.2673273086547852,
+      "grad_norm": 0.0,
+      "kl": 1.256190538406372,
       "learning_rate": 0.0001542857142857143,
-      "logps/chosen": -307.4684753417969,
-      "logps/rejected": -375.4805603027344,
-      "loss": 0.4081,
-      "rewards/chosen": -1.0870614051818848,
-      "rewards/margins": 3.670943021774292,
-      "rewards/rejected": -4.587001800537109,
+      "logps/chosen": -1522.9342041015625,
+      "logps/rejected": -1577.1060791015625,
+      "loss": 0.4463,
+      "rewards/chosen": -123.79159545898438,
+      "rewards/margins": 4.876855373382568,
+      "rewards/rejected": -128.13319396972656,
       "step": 40
     },
     {
       "epoch": 0.34,
-      "eval_kl": 1.7102612257003784,
-      "eval_logps/chosen": -316.2467346191406,
-      "eval_logps/rejected": -332.64117431640625,
-      "eval_loss": 0.38967660069465637,
-      "eval_rewards/chosen": -0.9805464744567871,
-      "eval_rewards/margins": 3.1516332626342773,
-      "eval_rewards/rejected": -3.9925851821899414,
-      "eval_runtime": 138.3084,
-      "eval_samples_per_second": 2.169,
-      "eval_steps_per_second": 0.542,
+      "eval_kl": 0.0,
+      "eval_logps/chosen": -1422.4874267578125,
+      "eval_logps/rejected": -1508.553955078125,
+      "eval_loss": 0.473336786031723,
+      "eval_rewards/chosen": -114.58648681640625,
+      "eval_rewards/margins": 10.318593978881836,
+      "eval_rewards/rejected": -124.07293701171875,
+      "eval_runtime": 140.0926,
+      "eval_samples_per_second": 2.141,
+      "eval_steps_per_second": 0.535,
       "step": 50
     },
     {
       "epoch": 0.41,
-      "grad_norm": 12.671459197998047,
-      "kl": 2.424811840057373,
-      "learning_rate": 0.00012714285714285714,
-      "logps/chosen": -321.6191101074219,
-      "logps/rejected": -360.3143310546875,
-      "loss": 0.4136,
-      "rewards/chosen": -0.7625396847724915,
-      "rewards/margins": 2.869812250137329,
-      "rewards/rejected": -3.8002796173095703,
+      "grad_norm": 3.258587184973294e-06,
+      "kl": 0.0,
+      "learning_rate": 0.00012571428571428572,
+      "logps/chosen": -1605.4840087890625,
+      "logps/rejected": -1731.5294189453125,
+      "loss": 0.4812,
+      "rewards/chosen": -133.641357421875,
+      "rewards/margins": 11.06646728515625,
+      "rewards/rejected": -144.48353576660156,
       "step": 60
     },
     {
       "epoch": 0.55,
-      "grad_norm": 6.753478050231934,
-      "kl": 5.016882419586182,
-      "learning_rate": 9.857142857142858e-05,
-      "logps/chosen": -314.6625671386719,
-      "logps/rejected": -389.6596374511719,
-      "loss": 0.4034,
-      "rewards/chosen": -1.544058084487915,
-      "rewards/margins": 5.419744968414307,
-      "rewards/rejected": -7.084101676940918,
+      "grad_norm": 0.0,
+      "kl": 0.0,
+      "learning_rate": 9.714285714285715e-05,
+      "logps/chosen": -2286.116455078125,
+      "logps/rejected": -2312.602294921875,
+      "loss": 0.4688,
+      "rewards/chosen": -200.4071044921875,
+      "rewards/margins": 2.4775140285491943,
+      "rewards/rejected": -200.7866973876953,
       "step": 80
     },
     {
       "epoch": 0.68,
-      "grad_norm": 0.0027679901104420424,
-      "kl": 0.5703033208847046,
-      "learning_rate": 7e-05,
-      "logps/chosen": -532.3701171875,
-      "logps/rejected": -549.9442749023438,
-      "loss": 0.4304,
-      "rewards/chosen": -21.35959243774414,
-      "rewards/margins": 2.940380096435547,
-      "rewards/rejected": -22.644678115844727,
+      "grad_norm": 0.0,
+      "kl": 0.0,
+      "learning_rate": 6.857142857142858e-05,
+      "logps/chosen": -2744.5634765625,
+      "logps/rejected": -2576.698974609375,
+      "loss": 0.4719,
+      "rewards/chosen": -246.3837890625,
+      "rewards/margins": -13.795512199401855,
+      "rewards/rejected": -228.7157440185547,
       "step": 100
     },
     {
       "epoch": 0.68,
       "eval_kl": 0.0,
-      "eval_logps/chosen": -1804.134765625,
-      "eval_logps/rejected": -1664.4483642578125,
-      "eval_loss": 0.4733425974845886,
-      "eval_rewards/chosen": -149.7693634033203,
-      "eval_rewards/margins": -9.425224304199219,
-      "eval_rewards/rejected": -137.1732940673828,
-      "eval_runtime": 138.3606,
-      "eval_samples_per_second": 2.168,
-      "eval_steps_per_second": 0.542,
+      "eval_logps/chosen": -2469.165771484375,
+      "eval_logps/rejected": -2225.683349609375,
+      "eval_loss": 0.47333332896232605,
+      "eval_rewards/chosen": -219.25433349609375,
+      "eval_rewards/margins": -26.235870361328125,
+      "eval_rewards/rejected": -195.78585815429688,
+      "eval_runtime": 139.9824,
+      "eval_samples_per_second": 2.143,
+      "eval_steps_per_second": 0.536,
       "step": 100
     }
   ],

checkpoint-100/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e6dcfcc81142dd54ec7ba9ddeb9a8bec23710b9520aa79594742249d14654995
+oid sha256:91658cc872c39fb432cb322f2450e9b0c776e35ec752fc4911f1c95a922ea14a
 size 5624